evolution of a landslide-induced sediment wave in the … permission to copy, contact...

For permission to copy, contact [email protected]� 2002 Geological Society of America

GSA Bulletin; August 2002; v. 114; no. 8; p. 1036-1048 13 figures; 1 table.

Evolution of a landslide-induced sediment wave in theNavarro River, California

Diane G. Sutherland*Maria Hansler BallSusan J. HiltonThomas E. LisleUSDA Forest Service, Pacific Southwest Research Station, 1700 Bayview Drive, Arcata, California 95521, USA

ABSTRACT

A streamside landslide delivered 60 000m3 of mixed-size sediment to the NavarroRiver, a sinuous gravel-bed channel (drain-age area � 535 km2), at the end of the an-nual high-runoff period in spring 1995. Thedeposit formed a 9-m-high dam that par-tially breached within several hours, but re-cessional flows entrained little material un-til the following high-runoff season. Thelandslide afforded the opportunity to mea-sure the evolution of a sediment wave fromits inception to near-obliteration and, par-ticularly, to test relative tendencies fortranslation and dispersion of a sedimentwave in a natural gravel-bed channel. Thisstudy represents a simple case: The waveoriginated from a single input, the preex-isting channel was relatively uniform, andresistant banks prevented adjustments inwidth. We surveyed channel topographyover a 1.5–4.5 km reach centered on thelandslide dam each year from 1995 to 1999,and we sampled bed material downstreamof the dam in 1995 and 1997. Landslide ma-terial was coarser than ambient bed mate-rial, but all sizes were mobilized by subse-quent peak flows. Abrasion of weatheredand fractured graywacke sandstone land-slide material was roughly an order of mag-nitude greater than the ambient rivergravel.

The sediment wave dispersed and mostlydisappeared within a few years with nomeasurable translation. Sediment filled thereservoir created by the eroding landslidedam until throughput of bed load was re-stored in 1998. The stationary wave cresteroded until in 1999 it was �1 m higher

*E-mail: [email protected].

than the preslide elevation. As the waveprofile flattened, its detectable leading edgeextended downstream from 620 m in 1995to �1600 m in 1997. Downstream advanceof the wave was associated with coarseningof bed material.

The sediment wave created a longitudi-nal disturbance in sediment transport. Byusing the dam as a reference datum of zerobed-load transport, we computed longitu-dinal variations in annual bed-load andsuspended-sediment transport rates in 100m increments downstream of the dam.These longitudinal variations were con-trolled by scour and fill of the bed and byabrasion of bed-load particles. Bed-loadtransport rates in the first and second yearsafter the landslide increased in the land-slide vicinity and then decreased down-stream as sediment deposited behind theadvancing leading edge of the wave. The lo-cation of peak bed-load transport rate ad-vanced from the first year (400 m) to thesecond (800 m).

We used a physically based, one-dimen-sional model (Cui et al., 2002b) to hindcastannual changes in transversely averagedbed elevation over the study reach. Agree-ment between measured and predicted bedelevations was very good. This result sup-ports our conclusion that, once emplaced,sediment waves in gravel-bed rivers tend todisperse, with little or no translation.

Keywords: bed-load waves, geomorpholo-gy, modelling, sediment transport.

INTRODUCTION

Understanding the movement of sedimentthrough channel systems is important to manygeomorphic problems. In the short term (100–

101 yr), sediment commonly carries the sig-nature of upstream disturbances in runoff anderosion to downstream channels and valleybottoms and their aquatic and riparian ecosys-tems. The arrival, duration, and intensity ofsediment impacts depend strongly on how ex-cess sediment is distributed and interacts withstored sediment once it reaches the channelsystem. At longer time scales (101–103 yr),sediment fluxes build and modify alluviallandforms.

Sediment loads in channels are commonlypunctuated in time and space by large inputsthat produce sediment ‘‘pulses’’ or ‘‘waves.’’Sediment waves are transient zones of sedi-ment accumulation in channels that evolve byinteractions between flow and sediment trans-port in the particular valley-bottom and chan-nel topography through which they propagate.We adopt the term ‘‘wave’’ to refer to a prop-agating disturbance in bed elevation and sed-iment properties that might translate and/ordisperse. Thus, a wave does not necessarilyconsist of only the original sediment of theinput, but can also incorporate and mix withpreexisting sediment in the river. The problemcan be stated rhetorically: How do river sys-tems digest large sediment inputs (Cui et al.,2002a)? Understanding how single sedimentwaves disperse and propagate should improveunderstanding of the routing of sediment frominputs distributed in a drainage network andthereby aid analyses of cumulative watershedeffects. Such understanding should also im-prove predictions of the fate of large volumesof sediment released from decommissioneddams. In this study, we focus on a wave ofbed material (sand and gravel) and neglectsediment transported in suspension.

A basic issue of the behavior of bed-mate-rial waves and their effect on stream channels

Geological Society of America Bulletin, August 2002

EVOLUTION OF A LANDSLIDE-INDUCED SEDIMENT WAVE

Figure 1. Models of sediment-wave evolu-tion. Lighter lines symbolize later stages ofevolution. (A) Sediment that is advectedfrom upstream and incorporated in thewave is apparent in the dispersing wave asdeposition upstream of the original wave.(B and C) In either of the translational cas-es, deposition immediately upstream of theinput point is not apparent.

Figure 2. Location of study site.is the relative tendencies for translation or dis-persion. For purposes of discussion, a purelydispersive wave is one whose trailing edgeand apex do not migrate downstream, al-though its center of mass may shift as its frontadvances (Fig. 1A). The backwater effect ofthe wave may cause sediment advected fromupstream to accumulate on the upstream limbof the wave. The result is a more symmetricspread of the wave about its point of origin.In a translational wave, both leading and trail-ing edges and the wave apex advance down-stream (Fig. 1B). A reasonable expectation isfor a wave to both translate and disperse tovarious degrees (Fig. 1C). The relativestrength of these tendencies is important forpredicting sediment impacts. Dispersion ofsediment inputs attenuates but prolongs im-pacts downstream, whereas translation pre-serves pronounced sequences of impact andrecovery as the wave migrates downstream. Ina natural channel, topography can force lon-gitudinal variations in deposition, making de-grees of dispersion and translation difficult todiscern. Monitoring bed elevation with littleknowledge of preexisting elevations at scat-tered points along the channel can lead to am-biguous interpretations of wave evolution, asnoted by Lisle et al., 2001.

Expectations of wave behavior have beeninfluenced by Gilbert’s (1917) landmark studyof sediment waves produced by hydraulicmining of placer deposits along tributaries ofthe American and Sacramento Rivers in Cal-ifornia. His conclusions (p. 30) regardingwave behavior emphasize translation: ‘‘The

downstream movement of the great body ofdebris is thus analogous to the downstreammovement of a great body of storm water, theapex of the flood traveling in the direction ofthe current.’’ Results of a numerical model ofBenda and Dunne (1997) predict translationand dispersion of sediment waves downstreamof sediment sources in a basin in the OregonCoast Range. Miller and Benda (2000) ob-served channel morphology and bed textureafter a high-magnitude flood in a steep chan-nel and interpreted their observations to indi-cate translation of a wave of debris-torrentmaterial. However, other studies of sedimentwaves in rivers show strong dispersion andweak or questionable translation (Roberts andChurch, 1986; Knighton, 1989; Pitlick, 1993;Turner, 1995; Madej and Ozaki, 1996). Recentnumerical models and laboratory experimentsindicate the dominance of dispersion in theevolution of sediment waves in gravel-bedchannels (Lisle et al., 1997; Dodd, 1998; Lisleet al., 2001; Cui et al., 2002a, 2002b). Wavetranslation is enhanced by low Froude num-bers (typical of low-gradient, sand-bed chan-nels) and wave material that is finer than am-bient streambed material (Meade, 1985; Lisleet al., 2001).

This paper reports detailed measurements ofthe evolution of a sediment wave in the Na-varro River in north coastal California fromits inception as a landslide through late stagesof evolution. The results provide an unambig-uous evaluation of the relative degrees of

translation and dispersion of an evolving sed-iment wave in a natural channel. In this case,widely graded input sediment is coarser thanthe ambient river gravel, but all sizes are mo-bile. We relate field observations to theory bytesting predictions of a one-dimensional nu-merical model of wave evolution (Cui et al.,2002b) that is based on first principles of con-servation of water mass and momentum andsediment, but also includes empirical valuesbased on field and laboratory data on graveltransport. Field and modeling results confirmthat the wave is exclusively dispersive.

STUDY SITE AND LANDSLIDE EVENT

The Navarro River is located in north coast-al California (Fig. 2) in a temperate maritimeclimate. Mean annual precipitation is�130cm, nearly all of which falls as rain from No-vember to March. Vegetation consists ofmixed hardwoods and conifers and prairiegrasslands in the headwaters, vineyards andorchards in the upper main stem, and second-growth forests along the lower main stem. Lo-cal lithology is the Coastal Range terrane ofthe Franciscan Formation, chiefly composedof highly sheared sandstone and siltstone aswell as some mudstone (Manson, 1984).Drainage area at the study reach is 535 km2.A U.S. Geological Survey gauging station(#11468000; drainage area� 785 km2) is lo-cated 25 km downstream of the study reach;it recorded an average 15-minute discharge of


SUTHERLAND et al.

Figure 3. View looking downstream at the landslide entering the Navarro River, March1995 (photograph provided by Julie Sowma-Bawcom, California Geological Survey). Ar-row points to remnants of the landslide dam, which provides the zero bed-load transportdatum. The sediment wave extended downstream to (X) in 1995. Asterisk indicates theapproximate location of BMA.

14.5 m3·s–1 for the period of record (1950–2000). Bankfull width and depth at the studyreach are�57 m and 5.1 m, respectively;bankfull discharge (Q2.33), calculated by pro-rating drainage area at the landslide site (68%)to the gauge, is 385 m3·s–1. Discharges with arecurrence interval of 2.33 yr correspond wellwith the depth at which we find many bankfullterraces.

On March 23, 1995, a translational-rota-tional landslide mobilized�100 000 m3 ofsoil and sheared and weathered bedrock anddelivered 60 000 m3 into a straight reach ofthe Navarro River (Fig. 3). The landslide oc-curred at the end of an extremely wet winteralong a preexisting weakness in the rock(Sowma-Bawcom, 1996). Before the land-slide, the hillslope supported a young forest oftan-oak (Lithocarpus densiflorus), mixed ev-ergreens (primarilyPseudotsuga menziesii,Quercus spp., andArbutus menziesii), Cea-nothus spp., and redwood (Sequoia semper-virens). The landslide buried a 120-m-longreach of river to a depth of 9.3 m. Discharge

(�130 m3·s–1) was subbankfull at the time, butthe ponded flow overtopped the landslide damand eroded a narrow channel over the next 12h, transporting and depositing sand, angulargravel, and woody debris over�620 m ofchannel downstream. Afterward, dischargewaned rapidly and remained low until the nextwinter. The residual landslide dam, whichcontained a sparse framework of large woodydebris, was 4.5 m high and created a pond thatextended 2.2 km upstream of the slide. Thepond (locally called Lake Navarro) filled withbed load over the course of the following 3yr.

Landslide material consisted of mudstones,sandstones, siltstones, and minor volcanic li-thologies. These rocks readily break alongpervasive shear planes and abrade to fine sandand silt. Landslide debris was coarser than thepreexisting bed material although sorting wassimilar (described later). Ambient gravel wascomposed of lithologies similar to those of thelandslide but was more rounded and compe-tent and included minor amounts of chert.

The study reach extends from 2.6 km up-stream to 1.9 km downstream of the landslide.Active floodplains are narrow, and the channelis moderately confined alternately by valleywalls and high (�5 m) terraces. Channelbanks are stable, consisting mostly of eitherbedrock or densely rooted fine alluvium. Bankerosion during the study period was minimaldespite local aggradation and the occurrenceof some large floods. The channel has a rela-tively simple, single-thread, bar-pool mor-phology with an average channel width of 44m, average pool spacing of 210 m (4.8 chan-nel widths), and a gradient of 0.0013. Largeroughness elements (e.g., large woody debris,outcrops, boulders) are uncommon, althoughwillows grow on bars within the active bed.

In summary, this landslide provided a rareopportunity to measure the evolution of awave of sediment in a natural channel. It wasfortunate for this purpose that the wave wasemplaced in a single event during waningstages of the high runoff season, which en-abled us to measure initial stages of wave evo-lution in detail. Channel morphology was rel-atively simple, uniform, and characteristic ofmany gravel-bed rivers. Erosion-resistantbanks eliminated channel widening as a sig-nificant adjustment to wave evolution. The in-put sediment was somewhat coarser than theambient bed material, but all particles weretransportable.

METHODS

Our strategy was to measure variations inchannel topography and bed texture annuallyover the entire reach affected topographicallyby the sediment wave until the wave wasmostly dissipated. Topographic data weretransversely averaged and compared to predic-tions of a one-dimensional model (Cui et al.,2002b). We measured particle-size distribu-tions of surface and subsurface bed materialto investigate sorting of wave sediment in re-lation to topographic changes. We experimen-tally measured abrasion rates of landslide ma-terial and ambient river gravel in order tomeasure the contribution of particle abrasionto downstream fining and sediment-wave evo-lution. We used topographic data and abrasionrates to analyze longitudinal variations intransport rates. In the following sections, wedescribe our methodology in detail.

Topographic Surveys

During the course of the investigation, thewave crest remained at the landslide dam andlogically divided the wave into an upstream



Figure 4. Channel topography of the downstream limb of the sediment wave. Numbers in red are longitudinal distances (in meters)from BMA. Dashed lines indicate edges of active bed used in calculating year-to-year change in bed elevation (meters above mean sealevel), and dashed red and black lines indicate edge of active bed through the landslide each year.

limb measured as negative distances frombench mark A (BMA) at the upstream edge ofthe dam and a downstream limb measured aspositive distances from BMA. In the first year,topographic surveys of the study reach ex-

tended from�300 m to 1200 m. The next year,the study reach was expanded to�2600 mupstream and downstream in anticipation ofthe spread of the sediment wave. We surveyedthe channel in greatest detail from 0 to 1200

m in 1995 and from 0 to 1860 m in 1996 and1997. From this we generated digital elevationmodels (DEMs) having a 1 m cell size. Weused cross-sectional surveys to expand cov-erage upstream and downstream in less detail


SUTHERLAND et al.

in 1995–1997 and to measure elevations in theentire study reach in 1998 and 1999. In theintensively surveyed reach, we used a totalstation electronic surveying instrument and aranging-laser with internal compass to surveybed elevations in closely spaced (�5 m),loosely gridded intervals. Point coverages,which are two-dimensional, computer-gener-ated representations of geographic data sets,were produced from Cartesian coordinate sur-vey points. Point coverages were transformedto triangulated irregular networks (TINs) byconnecting nearest-neighbor survey points asvertices to create a triangular surface networkand, finally, a 1 m (horizontal dimension)DEM (Hansler, 1999). We estimated surveyerror by comparing DEMs to independentlysurveyed cross sections at the same locations.The absolute value of the mean difference inelevation between a DEM point and a surveypoint was 0.04 m (�0.11 m) (Hansler, 1999).The fact that this value was no greater thanthe characteristic particle roughness of the bedin the vicinity of the landslide (D84 � 0.10 m)validates a detailed examination of two-di-mensional variations in bed elevation. Errorfor transversely averaged bed elevations waseven less (0.004 m) because of the large sam-ple size. These estimates of error apply equal-ly to cross-section surveys.

Longitudinal profiles of transversely aver-aged bed elevations were generated by pro-jecting elevations from the 1 m DEM gridonto a midchannel line that was generatedfrom the midpoint of the shortest distance be-tween lines defining left and right edges of theactive channel bed. Mean bed elevations werecomputed as the average of all points in a 20m interval and were plotted at the interval’smidpoint. The profiles we constructed by us-ing the entire bed represent more area on theoutside of a bend than on the inside. Longi-tudinal profiles computed in this way will dif-fer from profiles constructed by using crosssections, which weight the outside and insideof a bend equally. There is one major bend inthe study reach, and we consider it to be in-significant to the interpretation of our results.

Particle-Size Distributions of Bed Material

Exposure of most of the bed during the an-nual summer drought greatly facilitated bed-material sampling. We mapped and measuredsurface particle-size distributions from�200m to the downstream end of the intensivelysurveyed study reach in 1995–1997. We de-lineated patches of like particle-size distribu-tions that were categorized into unimodal dis-tributions of (1) sand and fine gravel, (2)

medium gravel, or (3) coarse gravel or verypoorly sorted or bimodal distributions of (4)fine or (5) coarse gravel. We conducted pebblecounts (Wolman, 1954) of no less than 100particles on a subsample of randomly selectedpatches of each distribution type in 1995 (31%by number) and 1996 (24%) and on everypatch in 1997. We averaged surface particle-size distributions over 50-m-long channel seg-ments to investigate longitudinal variations inparticle size and sorting. Pebble counts wererecorded in half-� intervals to�2� (4 mm),and smaller particles were combined into a�4mm category.

In the same reach, we collected 13 bulksamples of subsurface bed material in 1995and 10 in 1997. In 1995, we selected samplelocations at randomx-, y-coordinates to equal-ly represent each patch type by area coveringthe entire study reach. In 1997, we extractedone subsurface sample in each 200-m-longchannel segment downstream of the landslide.Correlations between surface and subsurfaceparticle-size distributions at the patch scale(Mosley and Tindale, 1985; Lisle and Madej,1992; Buffington and Montgomery, 1999) ledto the approximation that the reach-averagedsubsurface size distribution can be found un-der patches whose bed surface size distribu-tions equal the reach-averaged bed surfacesize distribution (Lisle and Madej, 1992). Byusing visual estimation and referring to sam-ple pebble-count data, we approximated theaverage surface particle-size distribution foreach segment, found a location on the bed thatcorresponded to the average surface distribu-tion, and sampled there. We used averages offour 1995 bed-material pits immediatelydownstream of the landslide to represent theparticle-size distribution of the initially depos-ited sediment-wave material. We sampled bedmaterial downstream of the sediment wave in1995 to represent the ambient bed-materialdistribution.

We extracted subsurface samples after re-moving the surface layer to a depth equal toD90. We air-dried and sieved samples at half-� increments on-site, except for a subsampleof �11.2 mm material that we sieved later ina laboratory down to a minimum sieve size of0.5 mm, which we assumed was the upperlimit of sizes that could be suspended. Indi-vidual subsurface samples ranged from 100 to425 kg. In most cases, this range met the cri-terion of Church et al. (1987) that the largestparticle in the sample has a mass of�1% ofthe total. However, near the sediment-inputsource, the presence of boulders required usto accept a 5% limit (Mosley and Tindale,1985).

Abrasion Rate

Breakdown and abrasion of landslide par-ticles were apparently rapid. In 1995, manylandslide particles on the stream bed surfacewere broken into small fragments, apparentlya result of the sand- and siltstone particleshaving disaggregated after being exposed tosurface conditions. Downstream, gravel fromthe landslide could easily be distinguishedfrom ambient river gravel by weathering col-oration, angularity, and obvious fractureplanes. In addition, while sampling bed ma-terial in 1995, we noticed that many of theparticles were very friable and broke easilyduring the sieving process.

We used an abrasion mill in the form of atractor tire to measure rates of abrasion of ini-tially deposited sediment-wave material col-lected within 50 m downstream of the slideand ambient gravel collected upstream. In theabrasion tests, we ran samples over distancestotaling 1000 m, which approximated themaximum apparent travel distance of land-slide material. For each sample, we placed 25kg of material with a representative size dis-tribution (D � 180 mm) in a 1-m-diametertractor tire, in which rubber vanes�1 cm inheight were glued to promote particle rolling.We held the tire upright, filled it up to thelower lip of the rim with the sample and water,sealed the opening, and rolled the tire a spec-ified distance at�2 m·s–1. All material wasthen removed from the tire, air-dried, sievedto separate the�0.7 mm fraction, and re-weighed. The same material minus the�0.7mm fraction was reloaded into the tire and runfor the next interval, and so on, until a totalof 1000 m was traversed. Results of prelimi-nary runs indicated that abrasion rates of thewave material varied with travel distance andwere significantly affected by wetting-dryingcycles, but such was not the case for ambientriver gravel. Therefore, we ran wave materialover 100 m intervals to replicate step distanc-es and ran ambient gravel over longerintervals.

It is highly uncertain whether we accuratelyreplicated natural abrasion rates on the riverbed. We selected tire speed to approximatemaximum bottom-water velocities, but parti-cle motion in the tire can be expected to differfrom that on the river bed. Nevertheless, ourmeasured abrasion rates were reasonable con-sidering the material (discussed later), andmodel predictions of sediment-wave evolutionwere not sensitive to moderate variations ininput abrasion rates.



Figure 5. Longitudinal profiles of transversely averaged bed elevation (Sutherland et al.,1998).

Figure 6. Annual hydrographs for the study site, October–April 1996–1999. Flow datafrom USGS gauging station #11468000 were adjusted by prorating discharge to the drain-age area at the landslide (68% of the area at the gauge).

CHANNEL TOPOGRAPHY

Initial Topography, 1995

The intensive 1995 survey of the down-stream limb of the wave (Fig. 4) documenteda morphology created by the landslide and thedam-break flood; subsequent flows were toosmall to cause significant sediment transportuntil the 1995–1996 winter. The width of thechannel that had incised through the landslide(channel distance from 30 to 150 m) was only10 m, or about one-fourth that of the preexist-ing channel. Debris eroded from the breachedlandslide was deposited as a downstream-thinning wedge that could be detected by an-gularity and lithology as far as 620 m, justupstream of a major left bend in the channel,although a few landslide clasts were found asfar as 900 m. The breach occurred on the leftside of the channel, but the channel down-stream was forced to the right by a large,wood-mantled bar (150–270 m) that was ap-parently deposited by a debris flow during latestages of the landslide event. The bar was ashigh as 2 m above the 1995 low-flow watersurface and approximately twice as high asbars elsewhere. A pool at the landslide siteand another at 250–300 m downstream thatwere evident on a 1993 aerial photographwere filled, but two new shallow troughs werecreated between new alternate bars from 400to 600 m.

The longitudinal profile of mean bed ele-vation shows a downstream-thinning sedimentwave with a maximum elevation (25.5 mabove mean sea level) that was 4.5 m abovethe estimated prelandslide mean bed elevationand extended downstream to 620 m (Fig. 5).In 1995, Lake Navarro had a water surfaceelevation of 23.1 m and extended 2200 m up-stream. A thin (0.1–0.5 m), fine-grained, low-density deposit blanketed the pond bed, and asandy delta with an estimated thickness of0.3–0.5 m prograded into Lake Navarro.

Water Year 1996

In water year 1996 (WY 1996; October 1,1995, to September 30, 1996), the highestflow (259 cm·s) was less than the mean annualflood (Fig. 6), and its duration was only 4 h.Nevertheless, there was substantial evolutionof the sediment wave. Erosion of the toe ofthe landslide increased channel width twofold(Fig. 4). The debris-flow bar at 150–270 mwas eroded away, resulting in a 3.5-m-deeppool, and a sandy reattachment bar was de-posited along the opposite bank in the lee ofthe landslide. More subtle changes down-

stream are revealed in a map of scour-and-filldepths (Fig. 7). Maximum changes in bed el-evation occurred in the first 250 m. Alternatebars and pools in the straight reach down-stream (300–600 m) shifted slightly. Fartherdownstream in the large bend and beyond,pools scoured as bars filled.

The 1996 longitudinal profile (Fig. 5)shows no translation of the wave in the firstyear. The apex remained at the toe of the land-slide but lost 2.7 m in elevation. The 1995 and1996 profiles crossed at�500 m, and a shortreach of fill (�0.4 m deep) extended down-stream to 650 m, indicating a slight extensionof the downstream limb of the wave. Beyondthat, the change in mean bed elevation did notexceed 0.3 m. There was a net loss of 16 000

m3 of sediment from the downstream limb. Onthe upstream limb, the lake level dropped 1.6m, and lake length decreased by 1000 m. Sed-imentation of mud, silt, sand, and fine gravelincreased mean bed elevation immediately up-stream of the dam as much as 0.8 m. From�1200 to �2000 m, sediment inundated al-ders, which commonly grow near summer-flow margins, indicating that the channel hadaggraded�0.5 m in WY 1996 compared toestimated preslide elevation.

Water Year 1997

Flows in the second winter of study (WY1997) were larger than those of the first (Fig.6). Four storm flows exceeded bankfull, in-


SUTHERLAND et al.

Figure 7. Depths of scour and fill in the downstream limb of the sediment wave.

cluding one that reached a peak 15-minutedischarge of 1150 m3·s–1 (�7 times bankfulldischarge) and sustained flows over bankfullfor 12 h. In total, bankfull flow was exceededfor 36 h in WY 1997.

This high flow resulted in further substan-tial evolution of the sediment wave (Figs. 4and 7). However, despite the greater flows ofthe second winter, net loss of material fromthe downstream limb (15 000 m3) was approx-imately equal to that resulting from the firstwinter. The river continued to erode the land-slide toe, regaining 80% of its preslide width.The sandy bar along the right bank at 200 mto 250 m aggraded�1 m. The left-bank al-ternate bar originating from the pool at 200 m

elongated and buried the left-bank troughdownstream at 500–600 m. Meanwhile, theright-bank trough adjacent to this bar (at400–600 m) elongated into the bend. Littlechange occurred in the bend, but downstreamthere was substantial deposition at the headof the large, left-bank bar originating at�900 m and in troughs (1000–1600 m) far-ther downstream.

Changes in the longitudinal profile in 1997continued the trends of the previous year (Fig.5). The apex of the wave was eroded another 1m but remained in place (at 70 m). The channeldownstream degraded a lesser amount (�0.5 m)to �600 m and aggraded (0.2–0.5 m) from 800to 1600 m. Downstream of 1600 m, the channel

scoured�0.5 m. The downstream limit of thewave is unclear. The advancing wedge can bediscerned as far as 1000 m, but bar-pool topog-raphy obscures systematic changes in mean bedelevation farther downstream.

On the upstream limb, Lake Navarrodropped 0.9 m and receded another 400 m inlength. Small deposits of well-rounded gravel(D50 � �10 mm) were found downstream ofthe dam, indicating that some bed load passedthrough Lake Navarro and over the landslidedam. Cross sections from�175 m to 0 mshowed scour of fine material that was depos-ited in WY 1995. Farther upstream to�2200m, average cross-section elevation increasedby �0.1 m.



Figure 8. Particle-size distributions of subsurface bulk samples of ambient bed materialand of initially deposited landslide material collected 50 m downstream of the landslidein summer 1995.

Water Years 1998 and 1999

Water years 1998 and 1999 had flood peaksgreater than those of WY 1996 but not nearlyas great as the large peak of WY 1997 (Fig.6). Changes in the longitudinal profile weresmall from WY 1998 and hardly detectablefrom WY 1999 (Fig. 5). Changes in the down-stream limb were minor, but slight aggradationcan be detected near the end of the study reach(1800 m), suggesting farther advance of thewave. On the upstream limb, a slightly down-stream-thickening wedge averaging�0.5 mthick was deposited from�1500 to�15 m inWY 1998. The wave crest eroded an addition-al 0.5 m and, as a result, no longer pondedwater.

The overall effect was to leave a broadhump in the profile over the entire studyreach. By 1998, the wave had no clearly de-fined apex and had lost most of its topograph-ic expression. The wave was primarily re-vealed by a transition from fine roundedgravel upstream of the landslide to coarse,subangular gravel downstream. Downstreamof �1500 m, the only evidence of the sedi-ment wave was a few subrounded cobbles.

In summary, our surveys document most ofthe evolution of a sediment wave from incep-tion to obliteration. Longitudinally, thesechanges indicate a sediment wave that is de-caying in amplitude as sediment is accretedupstream and is spread downstream. After 4yr, the wave evolved into a broad, low humpin the river profile and lost all obvious topo-graphic expression within 2 km of the land-slide. Its legacy was a discontinuity in bedtexture created by the lag of landslide parti-cles. We predict that the wave will disappearas lag particles break down further and mixwith ambient bed load.

At a smaller scale, bed-elevation changeswere manifested as deformation of bar-pooltopography. On the upstream limb of thewave, backwater deposition of fine sedimentfrom upstream tended to bury bar-pool topog-raphy. Downstream of the dam, large volumesof coarse landslide material obliterated pre-existing bars and pools and formed new onesthat later elongated. Farther downstream, barsand pools were preserved as the thinning sed-iment wedge extended. Bars were predomi-nantly filled, whereas pools were either filledor scoured.

PARTICLE-SIZE DISTRIBUTIONS OFBED MATERIAL

The landslide material delivered to thechannel was angular and very poorly sorted;

the material comprised clay to a few bouldersas large as 2 m in diameter. Deep erosion ofthe landslide toe in subsequent years indicates,however, that all sizes were transportable bythe available flows. Cobbles and coarse peb-bles were the predominant sediments initiallycovering the downstream limb of the wave;finer gravel and sand covered most of thechannel downstream of the wave.

For modeling purposes, we used the aver-age particle-size distribution of subsurfacematerial sampled in 1995 from four pits at200–400 m to represent the initially depositedsediment-wave material; we used the averagedistribution sampled in 1995 from three pitsdownstream of the leading edge of the wave(1000–1200 m) to represent ambient bed ma-terial. The initially deposited wave materialwas coarser than but sorted similarly to (D50

� 19.0 mm, whereDi equals the particle sizefiner than cumulative percentagei; � � 2.5,where� equals the standard deviation of theparticle-size distribution [phi scale]) ambientbed material (D50 � 6.7 mm;� � 2.4) (Fig.8).

In order to investigate longitudinal varia-tions in bed texture, we computedD16, D50,andD84 from surface particle-size distributionsthat were averaged over�50-m-long channelsegments in the downstream limb (Fig. 9). Foreach 50 m segment, averages were calculatedby weighting individual pebble counts bypatch area. Downstream fining in surface par-ticles is evident as decreases in D50 and D84

down to 700–1000 m, approximately wherethe wave merges texturally as well as topo-graphically with the ambient channel (Fig. 5).With time, the bed surface became coarser notonly over new sections inundated by the ad-vancing wave in 1996 and 1997, but also overthe original extent of the downstream limb

(620 m), as finer material was selectivelytransported from the supply-limited reachdownstream of the bed-load trap created bythe lake. Parallel variations in subsurface sizeindicated that bed armoring was pervasive butweak; the ratio of local values of surfaceD50

to subsurfaceD50 was generally�2. A slightdepression in surfaceD84 andD50 at �450 mbetween 1996 and 1997 was created by round-ed pebbles that were transported through thelake and over the remnants of the landslidedam and deposited among the reworked land-slide material during WY 1997. Slight in-creases in particle size were evident as fardownstream as 1500 m in 1997, and we foundcharacteristic landslide cobbles as far as 1800m. An increase in rounded pebbles over thedownstream limb was evident from pebble-count data at cross sections in later years (Fig.9). Finally, channel topography forced localfining of the bed surface, specifically at thedistal end of a large pool and point bar (�700m) and at the upstream end of a long pool(�1600 m) (Fig. 4). Pools centered at 1100 mand 1400 m did not show such a dip becausethey were flanked by large coarse bars.

ABRASION RATES

Rapid downstream fining and net loss oflandslide material was partly a result of par-ticle abrasion. Our abrasion-mill experimentof 100 m runs with cumulative travel distanc-es up to 1000 m showed a decrease in theSternberg abrasion coefficient (Sternberg,1875; Kodama, 1991;w in ,X–wM /M � ex 0

where Mx � mass of bed-load particles aftertravel distance X; M0 � mass of bed-load par-ticles at the starting point) fromw � 1.0 km–

1 at 100 m tow � 0.46 km–1 at 1000 m (Fig.10). The variation ofw with travel distance


SUTHERLAND et al.

Figure 9. Longitudinal variation in particle sizes D16, D50, and D84 of surface bed materialaveraged over 50-m-long segments of the downstream limb of the wave in 1995, 1996, and1997; and subsurface D50 measured in 1997.

fits a power relation well (w � 4.92x–0.344; r2

� 1.00). Shorter runs of 25 m using initiallydeposited wave material yielded values ofw

as high as 3 km–1. In contrast, ambient gravelshowed little variation inw with travel dis-tance, and thew value (0.13 km–1) remainedlower than w values of initially depositedwave material at the distal end of the evidentwave (1000 m). Initially, high abrasion rateswere likely caused by rapid wearing of thesharp edges and weathering rinds of regolithparticles. These trends suggest that the wavematerial will continue to wear faster than am-

bient gravel as it travels beyond the studyreach.

To our knowledge, the values ofw forwave material are the highest reported. Al-though this material is obviously weak, thehigh values ofw may be partly due to ourabrasion mill. The mean abrasion rate of theambient gravel (w � 0.13 km–1), which con-sists predominantly of graywacke and shale,provides a comparison with previously mea-sured values ofw. Maximum values for gray-wacke are 0.0039 km–1 (Shaw and Kellerhals,1982) and 0.27 km–1 for shale (Gies and Gell-

er, 1982). We conclude that our estimates ofabrasion rates may be high relative to otherlaboratory-determined rates, but by no morethan an order of magnitude.

LONGITUDINAL VARIATION IN BED-MATERIAL TRANSPORT

The sediment wave represents a disturbancein sediment transport throughout the affectedreach. To examine this disturbance, we ana-lyzed longitudinal variations in sedimenttransport for the first 2 yr after the landslide,assuming that our study reach included the en-tire affected reach. Trapping of bed load be-hind the landslide dam from 1995 to 1997 cre-ated a zero-transport datum from which tocompute longitudinal variations in annual sed-iment flux. Annual bulk volume of bed ma-terial transported from each 20 m section ofchannel to the next was computed from sedi-ment continuity,O � I � �S�A, whereO �output, I � input, �S � change in storage,andA � loss of bed load to abrasion.

Starting from the zero datum at the dam (I� 0), the output from one section is the inputfor the next section downstream. The fractionof the bed-load output that is abraded to sus-pended load is computed from the local abra-sion coefficient downstream of the dam (Fig.10). We assume that the abrasion rate down-stream of 1000 m equals that of the ambientgravel because scour and fill of the bed in thisreach involves mostly ambient material. Weonly examined longitudinal variations down-stream of the dam because of the lack of res-olution created by the wide spacing of surveysupstream of the dam. However, we used thecross sections upstream of the dam to estimateambient total bed material yield for WY 1997.This estimate represents a maximum becauseit includes fine material that would not havebeen deposited in the bed without the back-water effect of the dam. In this period, nearlyall material transported downstream of thedam originated from the landslide toe or thebed. Little of the fine sediment that passedthrough the pond was deposited on the down-stream limb of the sediment wave; thus thiscontribution can be disregarded.

In WY 1996 and WY 1997, bed materialtransport increased downstream of the dam,peaked, and then declined, whereas transportrates of suspended sediment derived fromabrasion of bed-load particles (not includingsediment transported through the pond) in-creased steadily downstream, surpassing bed-load transport rates in WY 1997 (Fig. 11).Peak rates were similar but shifted down-stream from 400 m to 800 m in the later year.



Figure 10. Variation of particle-abrasion rate with transport distance measured in anabrasion mill.

Figure 11. Longitudinal variation in annual transport rates of bed-load and suspendedsediment in the downstream limb of the sediment wave.

The downstream decrease in bed-materialtransport was caused by deposition behind theleading edge of the wave, as well as abrasionof bed material and conversion to suspendedsediment. In WY 1997, transport rates recov-ered farther downstream, perhaps manifestingoscillatory scour-and-fill behavior. In bothyears, the end of the surveyed reach was be-yond any observable expressions of the wave.

The merging of local particle size and gra-dient with ambient conditions at the leadingedge of the wave, as well as a lack of changein bed topography, indicates the point of re-establishment of bed-load transporting condi-tions uninfluenced by the wave. Therefore,calculated transport rates at the end of thestudy reach should approximate those associ-ated with ambient conditions in the channel,if it is assumed that we were truly beyond the

influence of the wave. The farthest-down-stream transport rate for WY 1997 (5800m3·yr–1) was half of that estimated by the rateof channel filling in the upstream limb (11 700m3·yr–1), which should approximate the am-bient bed-material load of the river for thatyear. However, as we have previously sug-gested in this paper, this discrepancy may bedue to the pond trapping material that wouldotherwise have been suspended and transport-ed downstream. Nevertheless, transport ratesat the end of the wave were lower in WY 1997than in WY 1996, despite the higher discharg-es in WY 1997. Surface particle size remainednearly constant between 1996 and 1997 (Fig.9), whereas gradient decreased (Fig. 5). Thischange may have locally decreased the mo-bility of upstream parts of the wave and thusmay account for the small difference in peak

flux rates between the years. However, ac-cording to our assumption, ambient transportconditions would have to gradually be re-established at some point downstream. Appar-ently, the resolution of our data was not suf-ficient to recreate this.

Changes in the wave profile can be ex-plained with a simple model. The wave causesa discontinuity in particle transport and dis-equilibrium longitudinally in transport rate,but at some point downstream of the wave,ambient transport capacity is satisfied by ero-sion of the bed upstream. At this point, thechannel is adjusted as it was upstream of thewave to an equal sediment load before thewave was formed. If the wave were composedof immobile material, ambient transport ca-pacity would be satisfied by scour of the bedstarting downstream of the wave front (Fig.12A). If the wave material was more mobile,more of the transport capacity would be sat-isfied by scour of the wave and less by scourof the bed downstream (Fig. 12B). If justenough of the wave were scoured to satisfycapacity, the wave would appear to decay inplace with no change in the bed downstream(Fig. 12C). Finally, if erosion of the wavemore than satisfied capacity, deposition wouldoccur downstream of the wave and the wavefront would advance (Fig. 12D). Evolvingwave profiles and corresponding longitudinalpatterns of bed-material transport can be usedto categorize wave behavior. Case C (of Fig.12) fits the evolutionary stage of the wave inWY 1996; case D (of Fig. 12) fits that of WY1997 (Figs. 5 and 11).

MODEL PREDICTIONS

We used a one-dimensional model designedfor routing sediment pulses in heterogeneous,armored, gravel-bed channels to reconstructthe 1995–1999 evolution of the wave andthereby add a theoretical basis for our obser-vations. The model is described elsewhere(Cui et al., 2002b) and is briefly summarizednext. The model allows for fining downstreamas a function of selective transport and attri-tion. The model incorporates the St. Venantshallow-water equations for water continuityand momentum with four governing equationsdescribing channel-bed evolution: (1) a formof the Exner equation for mass conservationof heterogeneous sediment, (2) a transfer func-tion to regulate the exchange of particle sizesbetween surface and subsurface layers of thebed (Toro-Escobar, 1995), (3) a bed-loadtransport function (the surface-based bed-loadtransport function of Parker [1990a, 1990b]),and (4) the Keulegan resistance relationship.


SUTHERLAND et al.

Figure 12. Evolution of idealized dispersive waves and associated longitudinal variationsin sediment transport. Cases: (A) Nonerosive wave. (B) Partially erosive wave. (C) Sta-tionary wave. (D) Advancing wave. Abrasion of wave material and conversion to sus-pended sediment are neglected.

TABLE 1. VALUES OF INPUT PARAMETERS USEDIN MODEL PREDICTIONS OF WAVE EVOLUTION

Initial channel characteristics Value

Original channel slope 0.00128 m/mChannel width 44 mSediment porosity 0.43Active-layer thickness (Lw) 0.8 mReach length 4.35 kmParticle-size distribution of

ambient bed materialD50 � 5.1 mm; � � 2.4

Average abrasion coefficientof ambient bed material

0.1

Wave characteristics Value

Average abrasion coefficientof initial wave material

0.3

Particle-size distribution ofinitial wave material

D50 � 15 mm; � � 2.5

Input location 2.5–3.35 km

To employ the model, we made some sim-plifying assumptions: (1) Nonuniformities inchannel width, slope, and particle size beforethe landslide were too small to significantlyaffect the evolution of the sediment wave. (2)There were no significant tributary inputs ofsediment or water. (3) Only gravel (D �2mm) is included as part of the wave. Abrasionproducts are lost to suspension.

Channel-specific data sets required as inputto the model (Table 1) are as follows: (1) Pre-disturbance channel characteristics includedtransversely averaged mean bed elevations at

channel distances, channel width, active layerthickness, reach length, bed porosity, ambientabrasion coefficient, and particle-size distri-butions of surface and subsurface sediments.(2) Initial sediment-wave characteristics com-prised transversely averaged mean bed eleva-tions at channel distances, abrasion coeffi-cient, and particle-size distributions ofsediment input. (3) Discharge records provid-ed the mean daily discharge from the down-stream gauging station adjusted for drainagearea.

In addition, the user specifies total calcula-tion time, time increment for calculations, thenumber of nodes, and the increment lengthover which to calculate. Inputs to the modelare detailed in Hansler (1999).

MODEL RESULTS

The numerical model reproduces the gen-eral forms of evolutionary stages of the mean-elevation profile reasonably well (Fig. 13).The modeled profile of the first year (1995) isa simplification of the sediment wave super-posed on the model-generated initial profile.Already there are bar-scale discrepancies ow-ing to local topographic forcing that persist inlater years. The model also overpredicts de-position upstream of the landslide because of

apparently inflated estimates of ambient bed-load transport. The model performs bestdownstream of the landslide, where generalagreement is excellent for every year. The sa-lient feature of modeled wave evolution isthat, in agreement with topographic data, peakaggradation remains at the initial entry pointof the landslide; thus the wave does not trans-late. An analysis of the sensitivity of resultsto uncertainties in input parameters is reportedelsewhere (Cui et al., 2002b). Predictions arerobust for reasonable ranges of input param-eters, and none show significant departuresfrom the general characterizations of predictedwave evolution described herein for the Na-varro River sediment wave.

DISCUSSION AND CONCLUSIONS

A large landslide entering a gravel-bed riverat the close of the annual high-runoff seasonoffered a rare opportunity to observe the evo-lution of a sediment wave from initial to finalstages. This case was relatively simple and of-fered a test of the relative tendencies of trans-lation and dispersion of sediment waves inheterogeneous, gravel-bed channels with bar-pool topography. The details of the case areas follows:

1. The wave entered the channel in a singleevent, and other than bed load carried into thereach from upstream, there were no significantadditional inputs to the study reach.

2. The preexisting channel was relativelyuniform. The channel was single-thread and,other than a single bend, was only slightly sin-uous. The only significant variations in width,depth, slope, or particle size were associatedwith bar-pool morphology.

3. Resistant banks restricted channel re-sponse to changes within the original bound-aries of the active channel.

4. Flow events in years after the landslide,some exceeding bankfull levels, were highenough to mobilize all particle sizes intro-duced by the landslide.

In the simplest case, the sediment inputwould have the same particle-size distributionas the ambient bed material. In the NavarroRiver case, the input material had a wider dis-tribution and a larger median size than ambi-ent bed material. Wide grading of input ma-terial would have promoted selective transportand thereby enhanced wave dispersion. How-ever, the coarseness of the landslide depositdid not prevent deep channel incision to near-ly the original bed elevation. Moreover, suchdifferences in particle size between input andambient probably characterize many large sed-iment inputs.



Figure 13. Comparisons between measured detrended mean bed elevations (mean gradientremoved) and bed elevations predicted by the numerical model. The modeled profile for1995 is the input wave form and thus is not a prediction.

Given the simplicity of these natural con-ditions, accurate reconstructing of the evolu-tion of longitudinal patterns of mean bed el-evation by a physically based, numericalmodel that incorporates interactions betweenunsteady flow, bed topography, and transportof heterogeneous bed load affirms the mainconclusion drawn from field observations. Thetendency for dispersion of sediment waves innatural gravel-bed channels is apparently sostrong that translation is undetectable, barringsome special conditions such as those de-scribed next.

The downstream shift in peak bed-load fluxin the downstream limb of the wave from WY1996 to WY 1997 did, however, manifest

translatory behavior of a property of the wave,albeit an outcome of purely dispersive redis-tribution of bed material. In waves where to-pographic data are insufficient to accuratelyidentify wave behavior, a more apparent trans-lation of bed-load transport rates could givethe impression of the dominance of transla-tion. Moreover, longitudinal variations in bedmobility may have as much or greater signif-icance to lotic ecosystems as variations in bedelevation or texture.

Tendencies for significant wave translationsmay lie outside characteristic fluvial condi-tions for gravel-bed channels. At one extreme,low Froude numbers, which are more typicalof sand-bed channels, can promote wave

translation, as can fine-grained wave materialin gravel-bed channels (Cui et al., 2002b;Lisleet al., 2001). At the other extreme, debrisflows, torrents or bed-load flows in steep,bouldery channels commonly form a pro-nounced, rapidly translating front of coarsematerial and woody debris that is followed bymore erodible, finer-grained material (Innes,1983; Costa, 1984; Iseya et al., 1990).

Greater channel nonuniformity than that ex-pressed in this relatively simple case may fa-vor even greater rates of dispersion. In gen-eral, greater nonuniformity would creategreater spatial variation in tractive forces. Thislikelihood would tend to promote greatertransverse and longitudinal variations in bed-load transport and cause particles in high-transport zones to race forward as others de-posited in peripheral zones are left behind.The agreement between a one-dimensionalmodel and a full-dimensional field case in thisstudy suggests that dispersion would have alsodominated in a still simpler channel. However,certain downstream patterns in transport ca-pacity at the reach scale may force down-stream-propagating sequences of aggradationand degradation that resemble translation of asediment wave. More specifically, a rivercould fill and scour in a downstream patternas a spreading wave advanced into a sequenceof sedimentation zones. Such large-scale var-iations in alluviation are well known (Church,1983; Grant et al., 1990), but system-scale in-teractions with sediment fluxes are poorly un-derstood. Recognition of the dispersive behav-ior of sediment waves in gravel-bed riversmay simplify sediment routing strategies, butgreater understanding of sediment routing innatural systems will require greater under-standing of sediment movement over spatialscales longer than those of individual sedi-ment waves such as the one reported here.

The dominance of dispersion in the evolu-tion of sediment waves in gravel-bed rivershas important implications for the manage-ment of lotic ecosystems. In the absence oftranslation of sediment waves, downstreamsedimentation tends to be pervasive ratherthan punctuated. Dispersion and overlappingof sediment waves from a number of sourcesand their interaction with large-scale channeltopography can make it difficult to attributean observed sediment effect to its source. Oneof the perceived drawbacks of the decommis-sioning of dams is that the potentially largevolumes of released sediment may cause largeimpacts downstream. However, among all thescenarios of the fate of such sediment releases,the one assuming maximum dispersion pre-dicts the least magnitude of impact at any


SUTHERLAND et al.

point downstream. Such general predictionsmay motivate applications of more compre-hensive models to predict sediment routingand channel response.

ACKNOWLEDGMENTS

Yantao Cui provided the numerical model andguided us in running it. Thembi Borras, Bryan Dus-sell, Gwen Erickson, Sam Flanagan, Sherman Gar-inger, Lex Rohn, Bonnie Smith, Stephen Tordoff,and Giovanni Vadurro provided assistance in fieldwork. The Lousiana-Pacific Corporation and Men-docino Redwoods Corporation provided logisticalsupport and access to the field site. Sam Flanagan,Alejandro Ramirez, and Lizzy Gilliam helped con-duct abrasion experiments. We gratefully acknowl-edge their assistance. Insightful reviews by MikeChurch, Peter Knuepfer, and Dave Montgomeryhelped to improve the paper.

REFERENCES CITED

Benda, L., and Dunne, T., 1997, Stochastic forcing of sed-iment routing and storage in channel networks: WaterResources Research, v. 33, p. 2865–2880.

Buffington, J.M., and Montgomery, D.R., 1999, A proce-dure for classifying textural facies in gravel-bed riv-ers: Water Resources Research, v. 35, p. 1903–1914.

Church, M., 1983, Pattern of instability in a wanderinggravel bed channel,in Collinson, J.D., and Lewin, J.,eds., Modern and ancient fluvial systems: Oxford,Blackwell Scientific Publications, p. 169–180.

Church, M., McLean, D.G., and Wolcott, J.F., 1987, Riverbed gravels: Sampling and analysis,in Bathurst, J.C.,Thorne, C.R., and Hey, R.D., eds., Sediment transportin gravel-bed rivers: Chichester, John Wiley and Sons,p. 43–88.

Costa, J.E., 1984, Physical geomorphology of debris flows,in Costa, J.E., and Fleischer, P.J., eds., Developmentsand applications or geomorphology: New York,Springer-Verlag, p. 268–317.

Cui, Y., Parker, G., Lisle, T.E., Gott, J., Hansler, M.E., Piz-zuto, J.E., Almendinger, N.E., and Reed, J.M., 2002a,Sediment pulses in mountain rivers: Part 1. Experi-ments: Water Resources Research (in press).

Cui, Y., Parker, G., Pizzuto, J.E., and Lisle, T.E., 2002b,Sediment pulses in mountain rivers: Part 2. Compar-ison between experiments and numerical predictions:Water Resources Research (in press).

Dodd, A.M., 1998, Modeling the movement of sediment

waves in a channel [M.S. thesis]: Arcata, California,Humboldt State University, 53 p.

Gies, R., and Geller, F.J., 1982, Pressure gradients and deg-radation at the hydrotransport of coarse washerywastes,in Eighth International Conference on the Hy-draulic Transport of Solids in Pipes: Johannesburg:Cranfield, England, BHRA Fluid Engineering, p. 227–240.

Gilbert, G.K., 1917, Hydraulic mining debris in the SierraNevada: U.S. Geological Survey Professional Paper105, 154 p.

Grant, G.E., Swanson, F.J., and Wolman, M.G., 1990, Pat-tern and origin of stepped-bed morphology in high-gradient streams, western Cascades, Oregon: Geolog-ical Society of America Bulletin, v. 102, p. 340–352.

Hansler, M.E., 1999, Sediment wave evolution and analysisof a one-dimensional sediment routing model, NavarroRiver, northwestern California [M.S. thesis]: Arcata,California, Humboldt State University, 128 p.

Innes, J.L., 1983, Debris flows: Progress in Physical Ge-ography, v.7, p. 467–501.

Iseya, F., Ikeda, H., Maita, H., and Kodama, Y., 1990, Flu-vial deposits in a torrential gravel-bed stream by ex-treme sediment supply: Sedimentary structure and de-positional mechanism, in Third InternationalWorkshop on Gravel-bed Rivers: Firenza, Italy: Cich-ester, England, John Wiley and Sons, 24 p.

Kodama, Y., 1991, Effect of abrasion on downstream grav-el-size reduction in the Watarse River, Japan: Fieldwork and laboratory experiment [Ph. D. dissertation]:Tsukuba, Japan, University of Tsukuba, 155 p.

Knighton, A.D., 1989, River adjustments to changes in sed-iment load: The effects of tin mining on the Ringa-rooma River, Tasmania, 1875–1984: Earth SurfaceProcesses and Landforms, v. 14, p. 333–359.

Lisle, T.E., and Madej, M.A., 1992, Spatial variation in ar-mouring in a channel with high sediment supply,inBilli, P., Hey, R.D., Thorne, C.R., and Tacconi, P.,eds., Dynamics of gravel-bed rivers: Chichester, JohnWiley and Sons, p. 277–291.

Lisle, T.E., Pizzuto, J.E., Ikeda, H., Iseya, F., and Kodama,Y., 1997, Evolution of a sediment wave in an exper-imental channel: Water Resources Research, v. 33,p. 1971–1981.

Lisle, T.E., Cui, Y., Parker, G., Pizzuto, J.E., and Dodd,A.M., 2001, The dominance of dispersion in the evo-lution of bed material waves in gravel bed rivers:Earth Surface Processes and Landforms, v. 26,p. 1409–1420.

Madej, M.A., and Ozaki, V., 1996, Channel response tosediment wave propagation and movement, RedwoodCreek, California, USA: Earth Surface Processes andLandforms, v. 21, p. 911–927.

Manson, M.W., 1984, Geology and geomorphic featuresrelating to landsliding, Navarro southeast (ColdSpring) 7.5 minute quadrangle, Mendocino County,California: Sacramento, California Department of

Conservation, Division of Mines and Geology,scale 1:24 000.

Meade, R.H., 1985, Wavelike movement of bedload sedi-ment, East Fork River, Wyoming: Environmental Ge-ology and Water Science, v. 7, p. 215–225.

Miller, D.J., and Benda, L.E., 2000, Effects of punctuatedsediment supply on valley-floor landforms and sedi-ment transport: Geological Society of America Bul-letin, v. 112, p. 1814–1824.

Mosley, M.P., and Tindale, D.S., 1985, Sediment variabilityand bed material sampling in gravel-bed rivers: EarthSurface Processes and Landforms, v. 10, p. 465–482.

Parker, G., 1990a, Surface-based bedload transport relationfor gravel rivers: Journal of Hydraulic Research, v. 28,p. 417–436.

Parker, G., 1990b, The ACRONYM series of PASCAL pro-grams for computing bedload transport in gravel riv-ers: Minneapolis, University of Minnesota, St. An-thony Falls Laboratory, External MemorandumM-220, 124 p.

Pitlick, J., 1993, Response and recovery of a subalpinestream following a catastrophic flood: Geological So-ciety of America Bulletin, v. 105, p. 657–670.

Roberts, R.G., and Church, M., 1986, The sediment budgetin severely disturbed watersheds, Queen CharlotteRanges, British Columbia: Canadian Journal of For-estry Research, v. 16, p. 1092–1106.

Shaw, J., and Kellerhals, R., 1982, The composition of re-cent alluvial gravels in Alberta river beds: Alberta Re-search Council Bulletin, v. 41, 151 p.

Sowma-Bawcom, J.A., 1996, Breached landslide dam onthe Navarro River, Mendocino County, California:California Geology, Sept.–Oct., p. 120–127.

Sutherland, D.G., Hansler, M.E., Hilton, S.J., and Lisle,T.E., 1998, Sediment wave evolution and channelmorphologic changes resulting from a landslide, Na-varro River, northwestern California: Geological So-ciety of America Abstracts with Programs, v. 30, no.7, p. A-360.

Sternberg, H., 1875, Untersuchungen Uber Langen undQuerprofil geschiebefuhrender Flusse: Berlin, Zeit-schrift fur Bauwesen, v. 25 p. 483–506.

Toro-Escobar, C.M., 1995, Differential transport and de-position of gravel-sand mixtures in an aggrading river,[M.S. thesis]: Minneapolis, University of Minnesota,121 p.

Turner, T.R., 1995, Geomorphic response of the MadisonRiver to point sediment loading at the Madison Slide,southwest Montana [M.S. thesis]: Bozeman, MontanaState University, 98 p.

Wolman, M.G., 1954, A method of sampling coarse riverbed material: Transactions, American GeophysicalUnion, v. 35, p. 951–956.

MANUSCRIPT RECEIVED BY THE SOCIETY JUNE 14, 2001REVISED MANUSCRIPT RECEIVED MARCH 14, 2002MANUSCRIPT ACCEPTEDMARCH 21, 2002

Printed in the USA

evolution of a landslide-induced sediment wave in the … permission to copy, contact...

Documents